arXiv subjects

Neil G. Marchant

Publications and source records attributed to Neil G. Marchant.

13 recordsLinked to original sources

arXiv2026

What Was That Again? Certified Robustness for Automatic Speech Recognition

Automatic Speech Recognition systems are notoriously both sensitive to adversarial and benign perturbations. While this has been repeatedly demonstrated using reference datasets, detecting such behaviors in deployed systems is incredibly challenging, due to the absence of oracle knowledge of the true transcription. We demonstrate that employing a certification-inspired mechanism can significantly decrease WER, increase recall, and decrease the Spearman correlation between confidence and WER. We achieve this through a dual-gate diagnostic pipeline: a Two-Sided Atomic Audit that accumulates statistical wealth to certify both token existence and adversarial exclusion, and a Rank-Based Tournament that selects the winning sequence. Our evaluations across four diverse architectures demonstrate up to a 55% relative reduction in Word Error Rate, while also providing granular word- and sentence-level certifications to enhance acoustic security.

cs.LG↗

arXiv2026

Room for Error: Large-Scale Simulation of Over-the-Air Acoustic Attacks

While voice control is rapidly becoming a ubiquitous vector of human-AI communication, the risks facing these systems remain poorly understood. This is, in part, a product of the difficulties in scaling strictly digital adversarial workflows to the physical world. These scale barriers have led the community to abstract away key acoustic factors relating to detectability and the influence of geometry on acoustics. These methodological and metrological shortcomings undermine our understanding of risk. We illuminate these issues through real-world testing, conceptual discussions, and a novel, high-throughput reality simulation framework. By testing over 8 million adversarial evaluations, we demonstrate that acoustic awareness yields relative Word Error Rate increases of up to 94.5\% under Whisper and wav2vec. We employ this framework to explore a formalize and operationalize a Dual-Form Signal to Noise Ratio to decouple source stealth from victim attack efficacy, resolving a crucial limitation in current works. This lays the groundwork for repeatable, verifiable research that embraces, rather than abstracts, the acoustic environment.

cs.SD↗

arXiv2026

Watermarks Attack Watermarks: Re-Watermarking as a Generic Removal Strategy

Watermarking combines an imperceptible change to an input image that will trigger a detector, to assert provenance and protect intellectual property. The literature has shown great interest in attacks on watermarking schemes: attackers are clearly motivated to steal copyrighted material or circumvent legislated deepfake protections. In this work, we make a simple-yet-powerful observation: that such attacks on watermarking-like watermarks themselves-seek an imperceptible change to an input image (now already watermarked) that will trigger a detector. This analogy comparing watermark attacks to watermarking itself is highly suggestive: that watermarks could be used to attack watermarks. Our first contribution validates this hypothesis. In rigorous experiments spanning 96 combinations of dataset, victim, and attack watermarks, we show that simply re-watermarking an already watermarked image reliably suppresses the original signal, without requiring gradients, surrogate models, or detection keys. Our second contribution is a simple classifier for detecting the presence and identity of an existing watermark in a given image. Surprisingly, experimental findings demonstrate outstanding overall accuracies 0.878-0.953. This result is of independent interest as a security vulnerability: research shows that method-specific attacks achieve substantially stronger removal than black-box attacks. Taken together, watermark identification combined with re-watermarking successfully reduces bit accuracy by at least 25% and up to 48%. Our work constitutes a cheap, generic, and highly effective attack pipeline, calling into question the reliability of current watermarking schemes to such a simple attack, as well as the value of existing sophisticated attacks.

cs.CR↗

arXiv2025

Where is the Watermark? Interpretable Watermark Detection at the Block Level

Recent advances in generative AI have enabled the creation of highly realistic digital content, raising concerns around authenticity, ownership, and misuse. While watermarking has become an increasingly important mechanism to trace and protect digital media, most existing image watermarking schemes operate as black boxes, producing global detection scores without offering any insight into how or where the watermark is present. This lack of transparency impacts user trust and makes it difficult to interpret the impact of tampering. In this paper, we present a post-hoc image watermarking method that combines localised embedding with region-level interpretability. Our approach embeds watermark signals in the discrete wavelet transform domain using a statistical block-wise strategy. This allows us to generate detection maps that reveal which regions of an image are likely watermarked or altered. We show that our method achieves strong robustness against common image transformations while remaining sensitive to semantic manipulations. At the same time, the watermark remains highly imperceptible. Compared to prior post-hoc methods, our approach offers more interpretable detection while retaining competitive robustness. For example, our watermarks are robust to cropping up to half the image.

cs.CV↗

arXiv2025

On the Bayes Inconsistency of Disagreement Discrepancy Surrogates

Deep neural networks often fail when deployed in real-world contexts due to distribution shift, a critical barrier to building safe and reliable systems. An emerging approach to address this problem relies on \emph{disagreement discrepancy} -- a measure of how the disagreement between two models changes under a shifting distribution. The process of maximizing this measure has seen applications in bounding error under shifts, testing for harmful shifts, and training more robust models. However, this optimization involves the non-differentiable zero-one loss, necessitating the use of practical surrogate losses. We prove that existing surrogates for disagreement discrepancy are not Bayes consistent, revealing a fundamental flaw: maximizing these surrogates can fail to maximize the true disagreement discrepancy. To address this, we introduce new theoretical results providing both upper and lower bounds on the optimality gap for such surrogates. Guided by this theory, we propose a novel disagreement loss that, when paired with cross-entropy, yields a provably consistent surrogate for disagreement discrepancy. Empirical evaluations across diverse benchmarks demonstrate that our method provides more accurate and robust estimates of disagreement discrepancy than existing approaches, particularly under challenging adversarial conditions.

cs.LG↗

arXiv2025

AdaptDel: Adaptable Deletion Rate Randomized Smoothing for Certified Robustness

We consider the problem of certified robustness for sequence classification against edit distance perturbations. Naturally occurring inputs of varying lengths (e.g., sentences in natural language processing tasks) present a challenge to current methods that employ fixed-rate deletion mechanisms and lead to suboptimal performance. To this end, we introduce AdaptDel methods with adaptable deletion rates that dynamically adjust based on input properties. We extend the theoretical framework of randomized smoothing to variable-rate deletion, ensuring sound certification with respect to edit distance. We achieve strong empirical results in natural language tasks, observing up to 30 orders of magnitude improvement to median cardinality of the certified region, over state-of-the-art certifications.

cs.CL↗

arXiv2024

Adaptive Data Analysis for Growing Data

Reuse of data in adaptive workflows poses challenges regarding overfitting and the statistical validity of results. Previous work has demonstrated that interacting with data via differentially private algorithms can mitigate overfitting, achieving worst-case generalization guarantees with asymptotically optimal data requirements. However, such past work assumes data is static and cannot accommodate situations where data grows over time. In this paper we address this gap, presenting the first generalization bounds for adaptive analysis on dynamic data. We allow the analyst to adaptively schedule their queries conditioned on the current size of the data, in addition to previous queries and responses. We also incorporate time-varying empirical accuracy bounds and mechanisms, allowing for tighter guarantees as data accumulates. In a batched query setting, the asymptotic data requirements of our bound grows with the square-root of the number of adaptive queries, matching prior works' improvement over data splitting for the static setting. We instantiate our bound for statistical queries with the clipped Gaussian mechanism, where it empirically outperforms baselines composed from static bounds.

cs.LG↗

arXiv2023

RS-Del: Edit Distance Robustness Certificates for Sequence Classifiers via Randomized Deletion

Randomized smoothing is a leading approach for constructing classifiers that are certifiably robust against adversarial examples. Existing work on randomized smoothing has focused on classifiers with continuous inputs, such as images, where $\ell_p$-norm bounded adversaries are commonly studied. However, there has been limited work for classifiers with discrete or variable-size inputs, such as for source code, which require different threat models and smoothing mechanisms. In this work, we adapt randomized smoothing for discrete sequence classifiers to provide certified robustness against edit distance-bounded adversaries. Our proposed smoothing mechanism randomized deletion (RS-Del) applies random deletion edits, which are (perhaps surprisingly) sufficient to confer robustness against adversarial deletion, insertion and substitution edits. Our proof of certification deviates from the established Neyman-Pearson approach, which is intractable in our setting, and is instead organized around longest common subsequences. We present a case study on malware detection--a binary classification problem on byte sequences where classifier evasion is a well-established threat model. When applied to the popular MalConv malware detection model, our smoothing mechanism RS-Del achieves a certified accuracy of 91% at an edit distance radius of 128 bytes.

cs.CR↗

arXiv2023

Bayesian Graphical Entity Resolution Using Exchangeable Random Partition Priors

Entity resolution (record linkage or deduplication) is the process of identifying and linking duplicate records in databases. In this paper, we propose a Bayesian graphical approach for entity resolution that links records to latent entities, where the prior representation on the linkage structure is exchangeable. First, we adopt a flexible and tractable set of priors for the linkage structure, which corresponds to a special class of random partition models. Second, we propose a more realistic distortion model for categorical/discrete record attributes, which corrects a logical inconsistency with the standard hit-miss model. Third, we incorporate hyperpriors to improve flexibility. Fourth, we employ a partially collapsed Gibbs sampler for inferential speedups. Using a selection of private and nonprivate data sets, we investigate the impact of our modeling contributions and compare our model with two alternative Bayesian models. In addition, we conduct a simulation study for household survey data, where we vary distortion, duplication rates and data set size. We find that our model performs more consistently than the alternatives across a variety of scenarios and typically achieves the highest entity resolution accuracy (F1 score). Open source software is available for our proposed methodology, and we provide a discussion regarding our work and future directions.

stat.ME↗

arXiv2021

Hard to Forget: Poisoning Attacks on Certified Machine Unlearning

The right to erasure requires removal of a user's information from data held by organizations, with rigorous interpretations extending to downstream products such as learned models. Retraining from scratch with the particular user's data omitted fully removes its influence on the resulting model, but comes with a high computational cost. Machine "unlearning" mitigates the cost incurred by full retraining: instead, models are updated incrementally, possibly only requiring retraining when approximation errors accumulate. Rapid progress has been made towards privacy guarantees on the indistinguishability of unlearned and retrained models, but current formalisms do not place practical bounds on computation. In this paper we demonstrate how an attacker can exploit this oversight, highlighting a novel attack surface introduced by machine unlearning. We consider an attacker aiming to increase the computational cost of data removal. We derive and empirically investigate a poisoning attack on certified machine unlearning where strategically designed training data triggers complete retraining when removed.

cs.LG↗

arXiv2021

Needle in a Haystack: Label-Efficient Evaluation under Extreme Class Imbalance

Important tasks like record linkage and extreme classification demonstrate extreme class imbalance, with 1 minority instance to every 1 million or more majority instances. Obtaining a sufficient sample of all classes, even just to achieve statistically-significant evaluation, is so challenging that most current approaches yield poor estimates or incur impractical cost. Where importance sampling has been levied against this challenge, restrictive constraints are placed on performance metrics, estimates do not come with appropriate guarantees, or evaluations cannot adapt to incoming labels. This paper develops a framework for online evaluation based on adaptive importance sampling. Given a target performance metric and model for $p(y|x)$, the framework adapts a distribution over items to label in order to maximize statistical precision. We establish strong consistency and a central limit theorem for the resulting performance estimates, and instantiate our framework with worked examples that leverage Dirichlet-tree models. Experiments demonstrate an average MSE superior to state-of-the-art on fixed label budgets.

cs.LG↗

arXiv2020

d-blink: Distributed End-to-End Bayesian Entity Resolution

Entity resolution (ER; also known as record linkage or de-duplication) is the process of merging noisy databases, often in the absence of unique identifiers. A major advancement in ER methodology has been the application of Bayesian generative models, which provide a natural framework for inferring latent entities with rigorous quantification of uncertainty. Despite these advantages, existing models are severely limited in practice, as standard inference algorithms scale quadratically in the number of records. While scaling can be managed by fitting the model on separate blocks of the data, such a naïve approach may induce significant error in the posterior. In this paper, we propose a principled model for scalable Bayesian ER, called "distributed Bayesian linkage" or d-blink, which jointly performs blocking and ER without compromising posterior correctness. Our approach relies on several key ideas, including: (i) an auxiliary variable representation that induces a partition of the entities and records into blocks; (ii) a method for constructing well-balanced blocks based on k-d trees; (iii) a distributed partially-collapsed Gibbs sampler with improved mixing; and (iv) fast algorithms for performing Gibbs updates. Empirical studies on six data sets---including a case study on the 2010 Decennial Census---demonstrate the scalability and effectiveness of our approach.

stat.CO↗

arXiv2017

In Search of an Entity Resolution OASIS: Optimal Asymptotic Sequential Importance Sampling

Entity resolution (ER) presents unique challenges for evaluation methodology. While crowdsourcing platforms acquire ground truth, sound approaches to sampling must drive labelling efforts. In ER, extreme class imbalance between matching and non-matching records can lead to enormous labelling requirements when seeking statistically consistent estimates for rigorous evaluation. This paper addresses this important challenge with the OASIS algorithm: a sampler and F-measure estimator for ER evaluation. OASIS draws samples from a (biased) instrumental distribution, chosen to ensure estimators with optimal asymptotic variance. As new labels are collected OASIS updates this instrumental distribution via a Bayesian latent variable model of the annotator oracle, to quickly focus on unlabelled items providing more information. We prove that resulting estimates of F-measure, precision, recall converge to the true population values. Thorough comparisons of sampling methods on a variety of ER datasets demonstrate significant labelling reductions of up to 83% without loss to estimate accuracy.

cs.LG↗